The consequences of reversible gill remodelling on ammonia excretion in goldfish (Carassius auratus)

Ammonia excretion in fish occurs predominantly at the gill with only a minor fraction (<10%) being excreted by the kidney (Maetz, 1972; Wood, 1993). Traditionally, ammonia (referring to the sum of NH₃ and NH⁺₄) excretion across the apical membrane of gill epithelial cells was thought to reflect variable contributions of passive diffusion of NH₃ through the lipid bilayer and electroneutral Na⁺/NH⁺₄ exchange (Maetz and Garcia Romeu, 1964; Maetz, 1973) (for reviews, see Walsh and Henry, 1991; Wright, 1995; Wilkie, 2002). However, in recent years, an alternative model has emerged in which the transepithelial diffusion of NH₃ and/or NH₄⁺ is facilitated by the presence of ammonia channels formed by Rh (Rhesus) proteins (Hung et al., 2007; Nawata et al., 2007; Nakada et al., 2007a; Nakada et al., 2007b; Shih et al., 2008; Hung et al., 2008; Tsui et al., 2009; Braun et al., 2009b) (see reviews by Weihrauch et al., 2009; Wright and Wood, 2009). Experiments on zebrafish (Danio rerio) using gene knockdown techniques (Braun et al., 2009a) have demonstrated that a significant fraction of ammonia excretion (∼50%) in larvae is probably occurring via the Rh proteins Rhag, Rhbg and Rhcg1. In pufferfish (Takifugu rubripes) (Nakada et al., 2007b) and adult zebrafish (Nakada et al., 2007a; Braun et al., 2009b), the branchial Rh proteins are arranged in a precise spatial sequence to allow the step-by-step movement of ammonia from the blood into the water across a discrete series of barriers. Thus, Rhcg1 and Rhcg2 are localized to the apical membranes of ionocytes and pavement cells, respectively whereas Rhbg is confined to the basolateral membranes of pavement cells. The initial step for ammonia entry into the lamellar epithelium is facilitated by Rhag which is localized to pillar cells (Nakada et al., 2007b) that form the vascular channels within the gill lamellae (Laurent and Dunel, 1980).

Regardless of the relative contributions of passive NH₃ diffusion, Na⁺/NH₄⁺ exchange and Rh protein-facilitated NH₃/NH₄⁺ movement, the net rate of ammonia excretion across the gill should be positively correlated with functional lamellar surface area, all else being equal. While changes in functional lamellar surface area can be mediated acutely via cardiovascular adjustments (Booth, 1979; Farrell et al., 1980), it is also clear that several fish species can chronically modify gill functional surface area by structural remodelling (for reviews, see Sollid and Nilsson, 2006; Nilsson, 2007). For example, crucian carp (Carassius carassius) (Sollid et al., 2003) and goldfish (Carassius auratus) (Sollid et al., 2005; Mitrovic et al., 2009; Mitrovic and Perry, 2009) experience a decrease in functional surface area as water temperature is lowered, which is reversible upon exposure of fish to acute hypoxia. In mangrove killifish (Kryptolebias marmoratus), functional surface area of the gill is reduced during aerial exposure (Ong et al., 2007). Such long-term adjustments of gill functional lamellar surface area reflects the presence or absence of a mass of cells in the interlamellar regions termed the interlamellar cell mass (ILCM) (Sollid et al., 2003). In goldfish and crucian carp, a lowering of water temperature triggers the formation of the ILCM which, in some cases, can lead to a near total covering of the gill lamellae (e.g. Sollid et al., 2003). The physiological benefit of this type of gill remodelling is probably derived from the declining energetic costs of osmoregulation because the functional loss of lamellar surface area will limit the passive loss of salts thereby reducing the need for active ionic uptake.

The physiological consequences of gill remodelling on the branchial transepithelial transfer of ions or gases are not well understood. However, from the few studies on goldfish, it would appear that fish may retain their ionoregulatory capabilities regardless of whether the lamellae are uncovered or covered because the ion transporting ionocytes migrate with the outer edge of the ILCM to maintain contact with the external environment (Mitrovic et al., 2009; Mitrovic and Perry, 2009). The present study was conceived to assess the possible consequences of gill remodelling in goldfish on the fundamental process of ammonia excretion. The toxicity of ammonia requires its effective removal yet the loss of surface area with remodelling would be expected to impede its excretion at the gill. Thus, it was hypothesized that the capacity for ammonia excretion is maintained in fish experiencing a loss of lamellar surface area owing to an increase in the expression of Rh proteins or their spatial redistribution.

All experiments were performed according to the University of Ottawa institutional guidelines which comply with those of the Canadian Council on Animal Care (CCAC). Goldfish, Carassius auratus Linnaeus 1758 (weighing 25.0±0.9 g, N=142) were obtained from Along's International (Mississauga, ON, Canada). They were first acclimated to 18°C in large fibreglass tanks filled with dechloraminated tap water. After at least 1 week, temperatures were either increased or decreased 1°C per day to achieve final acclimation temperatures of 7°C or 25°C. Animals were kept in the facility under a 12 h:12 h light:dark photoperiod and were fed at least once a day with commercial food pellets. Fish were kept under such conditions for at least 2 weeks prior to their use in experiments. For experiments requiring that food ration be controlled or withheld, fish were transferred to glass aquaria (20 litres) provided with water of either 7 or 25°C. To allow the long-term (7 day) acclimation to hypoxia, fish were transferred to, and held in smaller aquaria (7.5 litres). Twenty-four hours prior to any experiment, fish were moved from their holding tanks to individual boxes (approximately 600 ml water volume) provided with flowing aerated water at the appropriate acclimation temperature; fish were not fed.

Fish, acclimated to 7°C (N=16) or 25°C (N=18), were transferred to glass aquaria (20 litre) where they were fed a daily ration of 5% of their body mass for at least 6 days. This level of feeding was chosen because it was the maximal amount routinely consumed by the 7°C fish. Within each temperature group, six of the fish were fasted for 3 days. All fish were then transferred to individual boxes for a further period of 24 h when they were not fed. Thus for the fasted fish, food was withheld for a total of 4 days while the fed fish (like all experimental fish) were deprived of food for the 24-h period immediately preceding experimentation.

To monitor ammonia (NH₃+ NH₄⁺) excretion (J_net,amm), water flow to the boxes was stopped; water temperature was maintained by immersion of the boxes in a temperature controlled bath while water oxygenation (and mixing) was provided by continuous aeration. Water samples (2 ml) were removed after a 10 min mixing period and 60 min later, frozen (–20°C) and subsequently analyzed for ammonia levels. After a 10 min period of flushing the boxes with fresh water, the entire procedure was repeated twice. The final J_net,amm data represent the averages of the three consecutive 1-h trials. To correct for possible ammonia addition to the water from non-piscine sources, a seventh box without a fish was run as a blank. Because ammonia concentrations remained constant in the empty chambers, there was never any need to correct for exogenous ammonia production.

After the third flux period, fish were weighed after they were killed by anaesthetic overdose using a solution of benzocaine (ethyl-P-amino-benzoate; Sigma, St Louis, MO, USA).

Acclimated fish (7°C, N=18; 25°C, N=18) were weighed and transferred to individual boxes 24 h prior to further experimentation. Measurements of J_net,amm were conducted over two 1-h consecutive intervals as outlined above. Data from these two periods were averaged to provide a measure of resting J_net,amm rates. Fish were then removed, lightly anaesthetized using benzocaine (2.4×10^–4 mol l^–1) and injected intraperitoneally with 1.12 μmol g^–1 of 140 mmol l^–1 NH₄Cl (N=12 at each temperature); control fish (N=6 at each temperature) were injected with 1.12 μmol g^–1 of 140 mmol l^–1 NaCl. Fish were returned to their individual boxes and allowed to regain equilibrium and re-establish spontaneous swimming movements prior to beginning (<30 min) subsequent hourly measurements of J_net,amm over the next 3 h. Preliminary experiments in which fish were monitored for 24 h after ammonia loading demonstrated that increases in J_net,amm occurred only within the first 3 h. Water temperature and oxygenation were maintained as above. A separate group of fish (25°C; N=6) was injected with NH₄HCO₃ with J_net,amm being monitored for 3 h after injection.

Fish acclimated to 7°C (N=24) were weighed and transferred to individual boxes 24 h prior to further experimentation. After assessing J_net,amm over two consecutive 1-h flux intervals (data were averaged), individual fish (N=12) were removed, placed in a 7.5 litre aquarium and then chased with a small net for 5 min until exhausted. Exhausted fish were injected with NH₄Cl, returned to their boxes and J_net,amm was monitored hourly for a 3-h period. A group of control fish (N=12) was subjected to the same protocol except that they were not exercised.

Fish acclimated to 7°C were transferred in groups of 6 to 7.5 litre aquaria provided with flowing water (0.75 l min^–1) delivered by a water equilibration column. For hypoxia experiments (N=12), water P_O2 was reduced gradually to 10 mmHg over 24 h by bubbling the equilibration column with N₂. The P_O2 was maintained at 10 mmHg for a further 6 days during which time fish were fed a daily ration of 0.5% body mass (the hypoxic fish were unable to consume as much food as in Series 1). After 7 days, fish were removed, weighed and transferred to individual chambers provided with hypoxic water where they were left unfed for 24 h prior to experimentation. After measuring J_net,amm over two consecutive 1-h intervals, fish were removed and injected with NH₄Cl after which J_net,amm was assessed hourly over 3 h. Hypoxia was sustained during the measurements of J_net,amm by bubbling 1.5% O₂:98.5% N₂ through the water in the fish boxes using a two-channel gas mixer (Cameron Instruments Inc., Port Aransas, TX, USA). A separate group of control fish (N=12) was subjected to the same protocol except that they were maintained continually under normoxic conditions. A final group (N=12) was subjected to hypoxia for 7 days but then allowed to recover under normoxic conditions for 24 h prior to ammonia loading.

Fish acclimated to 7°C or 25°C (N=6 for each group) were removed from holding tanks, killed by a blow to the head and the gills dissected for analysis of Rh protein levels by western blotting or protein localization within the tissues by immunocytochemistry.

Total protein was prepared by homogenizing dissected goldfish gills on ice and subsequently incubating homogenate in protein extraction buffer [10 mmol l^–1 Tris (pH 7.4), 2% Triton X-100 with complete protease inhibitor cocktail (Roche, Mississauga, ON, Canada)]. Samples were centrifuged at 14,000 g for 20 min at 4°C and supernatants were stored at –20°C. The concentration of protein in each sample was determined using a bicinchoninic acid protein assay with BSA as the standard. After adding 2× Laemmli loading buffer (Bio-Rad, Mississauga, ON, Canada) and 2-mercaptoethanol (5% v/v), samples (75 μg per lane) were size-fractioned using 10% SDS-PAGE gel at 200 V. Following electrophoresis, proteins were transferred onto PVDF membranes (Bio-Rad) at 15 V for 1 h using a semi-dry transblot unit (Bio-Rad). After transfer, membranes were blocked in 5% milk–TBST (1× Tris-buffered saline with 0.05% Tween 20) for 3 h at room temperature, and then incubated with one of four primary antibodies diluted in 1% milk–TBST (1:2000) at room temperature overnight: rabbit polyclonal antibodies to the Rh proteins, Rhag, Rhbg and Rhcg2, were raised in Fugu (Nakada et al., 2007b) and a rabbit polyclonal antibody to Rhcg2 was raised against the zebrafish protein (Nakada et al., 2007a). The Fugu Rh protein antibodies were previously shown to cross react specifically with zebrafish proteins (Braun et al., 2009a). Following three 5-min washes in TBST, membranes were incubated with HRP-conjugated anti-rabbit IgG secondary antibody (Bio-Rad) diluted in 1% milk–TBST (1:5000) for 1 h at room temperature. After incubation, membranes were washed four times for 10 min each in TBST. Bands were visualized using chemiluminescence substrates (Pico or Femto; Pierce, Rockford, IL, USA). Images were acquired by ChemiDoc equipped with Quantity-One 1-D analyzer software (Bio-Rad).

After probing with one of the antibodies against Rh proteins, membranes were stripped using Re-Blot Plus (Chemicon International Inc., Millipore, Billerica, MA, USA) and re-probed with a mouse monoclonal antibody raised against Tetrahymena thermophilia α-tubulin (12G10; Developmental Studies Hybridoma Bank, University of Iowa) with the following modification. Membranes were blocked with 5% milk–TBST for 1 h, incubated with primary antibody (1:1000) for 2 h and incubated with HRP-conjugated anti-mouse IgG antibody (1:5000) for 1 h. The density of protein bands was determined using ImageJ software (NIH). To control for differences in protein loading, the densities of the Rh protein bands were normalized to α-tubulin band densities.

Dissected gill arches were fixed overnight at 4°C in 4% paraformaldehyde (PFA) prepared in PBS. The samples were then cryoprotected by sequentially incubating in 15% and 30% sucrose solutions (in PBS). The gills were embedded in OCT cryo-sectioning medium (VWR, Mississauga, ON, Canada), incubated for 1 h at room temperature and sectioned at 10 μm thickness using a cryostat (Leica CM 1850 Laboratories Eq., Nussloch, Germany). Sections were collected on SuperFrost Plus glass slides (Fisher Scientific, Ottawa, ON, Canada).

Sections were baked at 42°C for 15 min for simplified heat-induced epitope retrieval. Sections were then rehydrated in PBS for 5 min, and blocked with 5% goat serum in PBST (0.5% Triton X-100 in PBS) for 1 h at room temperature. To remove residual Triton X-100, slides were washed three times for 5 min with PBS and incubated with one of the following primary antibodies (α5, Rhag, Rhbg, Rhcg1 or Rhcg2) overnight at room temperature. α5, a mouse monoclonal antibody raised against the α-1 subunit of chicken Na⁺/K⁺-ATPase, was obtained from Developmental Studies Hybridoma Bank (University of Iowa); all antibodies were diluted at 1:100 in PBS. As a negative control for background staining, some slides were incubated with PBS lacking primary antibody. All slides were washed three times for 5 min following the incubation with primary antibody, and then incubated for 1 h in the dark at room temperature with Alexa-Fluor-conjugated secondary antibodies (1:100 in PBS) for immunofluorescence detection (Alexa-Fluor-546-conjugated goat anti-mouse IgG and Alexa-Fluor-488-conjugated anti-rabbit IgG; Molecular Probes, Burlington, ON, Canada). After three 5-min washes with PBS, DAPI-containing mounting medium (Vector Labs, Burlingame, MA, USA) was applied to the slides, to stain the nuclei, and coverslips were placed on top. Slides were observed using an Axiophot (Zeiss, Munich, Germany) microscope with an Olympus DP70 digital microscope camera and Image Pro Plus software, Version 6.0 (Media Cybernetics, Bethesda, MD, USA).

Water samples were analyzed for ammonia levels using the colorimetric assay of Verdouw (Verdouw et al., 1978) modified for microplate use. Absorbance at 650 nm was determined using a plate reader (Devices Spectramax Plus 384, Concord, ON, Canada).

Water P_O2 was measured using a Clark type O₂ electrode connected to an O₂ meter (Cameron Instruments, Port Aransas, TX, USA). The analogue P_O2 data were acquired continually using Biopac (Harvard Instruments, Montreal, QC, Canada) data acquisition hardware and software (AcqKnowledge, version 3.7.3). The O₂ electrode was calibrated using a zero solution (2% sodium sulphite) and air-equilibrated water maintained at the appropriate acclimation temperature.

The first right gill arch of each fish was removed and fixed in 4% paraformaldehyde (PFA) until it was ready for sectioning. The gills were placed in 30% sucrose (w/v) solution 24 h before sectioning to dehydrate and cryoprotect the tissue. Prior to sectioning they were placed in OCT Cryomatrix (Thermo Scientific, Inc., Ottawa, ON, Canada) for at least 1 h before being frozen at –30°C. The tissue sample was then sliced into 12–14 μm thick sections using a Leica CM3050 (Leica Microsystems, Inc., Richmond Hill, ON, Canada) cryostat and stained with Haematoxylin (Harris Hematoxylin with glacial acetic acid; Poly Scientific, Bay Shore, NY, USA) to visualize the tissue. Using light microscopy, photos of six gill filaments were taken randomly per section for six sections per fish (36 photos per fish). An Axiophot (Zeiss, Germany) microscope with an Olympus DP70 digital camera were used to obtain the images and the relative cross sectional area of the ILCM was determined using Image Pro version 6.0 and Image-J software (http://rsbweb.nih.gov/ij/index.html).

All data are presented as means ± 1 s.e.m. Ammonia excretion data were analyzed either by two-way ANOVA (Series 1), repeated measures two-way ANOVA (Series 2–4) or unpaired Student's t-test (Series 5). If appropriate, post hoc multiple comparison tests were performed after ANOVA to identify the significantly different data points (Bonferroni t-test). All statistical analyses were performed using a commercial software package (SigmaStat version 3.5; SPSS Inc.).

At both acclimation temperatures, there was a pronounced effect of feeding history on J_net,amm with the fed fish exhibiting markedly higher excretion rates (P<0.001; Fig. 1). Regardless of nutritional status, however, acclimation temperature did not significantly influence the rates of J_net,amm (P=0.091; Fig. 1); over the 18°C temperature range, the Q₁₀ values for fed and starved fish were only 1.07 and 1.20, respectively.

Ammonia excretion was unaffected by handling and physical manipulation because intraperitoneal injections of NaCl at both acclimation temperatures (Fig. 2A) was without effect (P=0.70). Thus, the significant increases (P<0.001) in J_net,amm following the injection of NH₄Cl (Fig. 2B) were a specific consequence of internal ammonia loading. The clearance of the ammonia load was markedly enhanced (P<0.001) in the fish acclimated to 25°C (Fig. 2B). At 25°C, the total amount of ammonia excreted over 3 h was 1.00±0.20 μmol g^–1 or roughly double the amount cleared at 7°C (0.45±0.08 μmol g^–1).

Injecting fish with equimolar quantities of NH₄HCO₃ (data not shown) caused J_net,amm to increase in an identical manner to that of NH₄Cl injection (e.g. J_net,amm was increased to 1.13±0.1 μmol g^–1 h^–1 during the first measurement period following NH₄HCO₃ injection in comparison with 1.16±0.1 μmol g^–1 h^–1 following NH₄Cl injection).

Prior chasing until exhaustion of fish acclimated to 7°C resulted in greater J_net,amm in the first 1 h of measurement following the injection of NH₄Cl (Fig. 3); J_net,amm increased from 0.34 to 0.67 μmol g^–1 h^–1 in the control fish and from 0.34 to 1.02 μmol g^–1 h^–1 in the exercised fish.

The size of the ILCM was unaffected by the chasing protocol; the extent of ILCM coverage (as a percentage of total interlamellar area) was 83.2±5.0 (N=6) and 75.6±3.3% (N=6) in the control and exercised fish, respectively (data not shown).

Fish exposed to hypoxia for 1 week exhibited a significantly greater rate of ammonia clearance after ammonia loading when assessed under continuing conditions of hypoxia (Fig. 4). To eliminate the potential contribution of hypoxia itself to ammonia clearance, the experiment was repeated on fish that were returned to normoxia for 24 h. In these fish, the rate of ammonia clearance was also higher than in control fish (∼twofold higher; Fig. 4) and was similar to the rate of ammonia clearance observed in the fish acclimated to 25°C. There was no difference in the rate of ammonia clearance between the previously hypoxic fish assessed under normoxic or hypoxic conditions (P=0.084; Fig. 4).

The relative expression levels in the gill of the four Rh proteins, as assessed by western blotting, remained constant regardless of acclimation temperature (Fig. 5). However, the pattern of their distribution within the gill appeared to be modified by the gill remodelling associated with temperature change. As previously shown in pufferfish (Nakada et al., 2007b) and zebrafish (Braun et al., 2009b), Rhag was localized to the pillar cell membranes (Fig. 6) presumably to act as the initial exit pathway for ammonia from the lamellar blood channels to the lamellar epithelial cells. Additionally, some lamellar epithelial cells exhibited Rhag immunoreactivity; these cells did not appear to be enriched with Na⁺/K⁺-ATPase. A subset of the Na⁺/K⁺-ATPase-enriched cells on the filament epithelium exhibited Rhag immunoreactivity (Fig. 6C). With the profound remodelling of the gill in fish acclimated to 7°C, the Na⁺/K⁺-ATPase-enriched cells were redistributed to the outer edge of the ILCM with most of them showing Rhag immunoreactivity (Fig. 6D).

In fish acclimated to 25°C, Rhbg was found at the basolateral membranes of lamellar epithelial cells, some of which were enriched with Na⁺/K⁺-ATPase (Fig. 7B). With the appearance of the ILCM at 7°C, the lamellar localization of Rhbg was maintained but there also appeared to be distinct localization to the Na⁺/K⁺-ATPase-enriched cells of the outer edge of the ILCM as well as a more diffuse distribution within the cell mass itself (Fig. 7C).

Rhcg1 was distributed along the lamellar epithelium where it appeared to be specifically localized to the apical membranes of all cells regardless of whether or not they were enriched with Na⁺/K⁺-ATPase (Fig. 8B,C). With the redistribution of the Na⁺/K⁺-ATPase-enriched cells to the outer edge of the ILCM at 7°C, the pattern of lamellar staining was retained, yet there was less evidence of colocalization of Rhcg1 and Na⁺/K⁺-ATPase (e.g. Fig. 8D). Occasionally, however, higher magnification images did reveal some Rhcg1 immunoreactivity in the Na⁺/K⁺-ATPase-enriched cells of the ILCM outer edge (Fig. 8E). The distribution pattern of Rhcg2 within the gill resembled that of Rhcg1 (Fig. 9).

The results of the present study demonstrate that, despite the profound structural remodelling of the gill that occurs during acclimation to cold water (7°C), goldfish are able to sustain ammonia excretion. Thus, although the results have demonstrated that the ILCM presents a barrier for ammonia excretion, one or more compensatory factors are apparently contributing to permit normal rates of excretion. The present findings demonstrate that the capacity of goldfish to excrete ammonia in the face of a diminished lamellar surface area may reflect, at least in part, a spatial redistribution of the branchial ammonia-transporting Rh proteins. Clearly, the continued ability of goldfish to excrete ammonia in the face of the impediment posed by the ILCM is crucial to the success of thermally induced gill remodelling as a strategy to conserve energy.

In addition to modifying surface area, acclimation of goldfish to differing temperatures is expected to influence a myriad of other factors that could potentially influence the rates of branchial ammonia excretion. Such temperature-dependent processes include metabolic rate, food consumption, nitrogen metabolism, ventilation and cardiac output. Thus, a challenge in the present study was to distinguish between the specific effects of reduced surface area and these other temperature-dependent changes on rates of ammonia excretion during thermally induced gill remodelling. This was achieved by comparing fed and fasted fish and by assessing ammonia clearance rates after loading with equimolar quantities of NH₄Cl. In addition, ammonia excretion was assessed in fish maintained at a constant temperature (7°C) while manipulating convection (following exhaustive exercise) or functional lamellar surface area (following exposure to hypoxia). Although increased perfusion of more distal lamellae during intense exercise might also be expected to increase functional surface area, the presence of an ILCM in the fish acclimated to cold water would presumably lessen the impact of such a response because the distal lamellae are covered. The results of these experiments revealed that the ability to excrete the exogenous ammonia load in the cold-acclimated fish was probably being constrained both by reduced rates of gill ventilation and perfusion as well as by a decrease in functional lamellar functional surface area.

The results of previous studies have demonstrated that the effects of temperature on ammonia excretion are complex and subject to the confounding influences of temperature on feeding, metabolic rate and the activities of several key catabolic proteins (Wood, 2001; Ballantyne, 2001). However, in studies on rainbow trout that have simultaneously measured feeding, nitrogen excretion and protein synthesis, it is clear that ammonia excretion decreases with falling temperature owing to the combined effects of decreased food consumption, protein synthesis and overall metabolic rate (Wood, 2001). In the present study, the effects of temperature acclimation on ammonia excretion were assessed in fed and fasted fish. The fed normoxic fish were given a common food ration of 5% body mass per day which corresponded to near maximal rates of food consumption for the 7°C fish but probably meant that the 25°C fish were being subjected to dietary restriction relative to the fish in cold water. Thus, as a proportion of overall metabolic rate, it is likely that the fish acclimated to 25°C were receiving fewer calories. We considered this to be less problematic than comparing fish with differing absolute caloric intakes. Under this condition of matched dietary food intake there was no significant effect of the 18°C temperature change on rates of ammonia excretion; the Q₁₀ value was 1.07. This result is in marked contrast to a Q₁₀ value of 4.0 reported for goldfish acutely transferred from 16 to 6°C (Maetz, 1972); fish were not fed for at least 24 h in those experiments. Because fish in that same study injected with NH₄Cl exhibited a lower average Q₁₀ of 1.9 (compared with 4.0), it was concluded that a significant component of the reduced ammonia excretion with falling temperature was caused by a reduced rate of nitrogen metabolism (Maetz, 1972). The fasted fish in the present study exhibited a Q₁₀ of 1.20 which is low relative to other studies (Wood, 2001) and surprising given that these fish were also experiencing a reduction in functional lamellar surface area, a possible decrease in the rate of endogenous ammonia production and reduced rates of ventilation and perfusion. Although the results of these experiments would appear to indicate that the formation of an ILCM in the cold-acclimated fish does not impede ammonia excretion, this conclusion should be treated with caution (see below) owing to the confounding effects of temperature on metabolism and physiological function. Thus, additional experiments were performed to control for ammonia production and to assess the potential roles of hyperventilation and perfusion on rates of ammonia excretion.

To control for altered rates of ammonia production, ammonia excretion was monitored in fish that were injected intraperitoneally with NH₄Cl. Because loading with NH₄Cl is likely to cause metabolic acidosis (Cameron and Kormanik, 1982), subsequent increases in ammonia excretion could reflect the need to excrete the exogenous ammonia load (via any route) as well as a specific increase in Na⁺/NH⁺₄ exchange triggered by the metabolic acidosis. To ensure that fish were responding to the ammonia load rather than metabolic acidosis, a separate series of experiments was performed in which fish were injected with an equimolar quantity of NH₄HCO₃, a protocol expected to cause metabolic alkalosis. Because the increases in ammonia excretion following loading were identical regardless of whether NH₄Cl or NH₄HCO₃ was used, it is clear that fish were responding predominantly to the increasing levels of ammonia rather than a changing blood acid–base status.

The results clearly demonstrated that the fish acclimated to cold water were unable to clear the exogenous ammonia load as effectively as the fish acclimated to warm water (Fig. 2B). The warm water fish were able to excrete 90% of the injected ammonia load within the 3 h measurement period compared with only 40% in the fish acclimated to cold water. Although these data are consistent with the ILCM acting as a barrier to ammonia excretion, it is also possible that the slower clearance of ammonia in the cold-acclimated fish reflects reduced ventilation volume and/or cardiac output.

In an attempt to ascertain whether the decreased levels of ventilation or cardiac output in the fish acclimated to cold water was limiting ammonia excretion, fish acclimated to 7°C were exercised to exhaustion immediately prior to ammonia loading. The underlying assumption was that ventilation and cardiac output would be elevated for at least a portion of the 3 h post-injection measurement period. Although exhaustive exercise also is known to markedly increase the production of ammonia in white muscle, this additional pool contributes little to branchial ammonia excretion (at least in rainbow trout over a 4 h post-exercise interval) (Wood, 1988) because it is largely contained within the intracellular compartment (Wood, 2001). Thus, we believe that the increased clearance of ammonia in the fish subjected to exhaustive exercise reflects increasing ventilation and cardiac output in addition to other factors that might be expected to enhance excretion (e.g. increasing blood pressure, increasing circulating catecholamines) (Farrell et al., 1979; Perry and Bernier, 1999). It is interesting that similar rates of ammonia clearance were achieved in the exercised 7°C fish as in the warm water control fish despite the presence of an ILCM in the former (note that the ILCM was unaffected by the exercise treatment). It is unlikely that non-branchial ammonia excretion was significantly contributing to ammonia clearance because Maetz (Maetz, 1972) reported that the goldfish kidney could account for no more than 8% of whole body ammonia excretion after an ammonia load.

To more specifically assess the impact of functional lamellar surface area on ammonia excretion, goldfish acclimated to 7°C were exposed to hypoxia (10 mmHg) for 7 days, a protocol that is known to cause the ILCM to disappear (Mitrovic et al., 2009). This allowed us to compare the capacity of fish at a constant temperature, but with widely differing functional lamellar surface areas, to excrete an exogenous ammonia load. The increased rates of ammonia clearance that were found to be associated with prior hypoxia exposure provide evidence that the presence of the ILCM impedes the transfer of ammonia across the gill. This finding is consistent with the results of Mitrovic et al. (Mitrovic et al., 2009) who observed an increase in the branchial efflux of polyethylene glycol (PEG 4000) in cold-acclimated goldfish experiencing a loss of the ILCM following hypoxia exposure. However, in that study (Mitrovic et al., 2009), the augmentation in PEG flux was observed only when fish were assessed under conditions of normoxia.

Given the obvious impairment of ammonia clearance associated with the presence of an ILCM, the absence of an effect of cold acclimation on the excretion of endogenous ammonia in fed or fasted fish (see above) suggests that fish were compensating for the lack of functional lamellar surface area. In this study, we tested the hypothesis that one such compensatory mechanism was an increase in, or redistribution of, the ammonia-transporting Rh proteins.

In pufferfish (Takifugu rubripes) (Nakada et al., 2007b) and zebrafish (Danio rerio) (Braun et al., 2009b), the Rh proteins are arranged spatially within the gill to provide a stepwise pathway for the excretion of ammonia from the blood to water. The movement of ammonia from the blood channels into the lamellar epithelium is believed to be facilitated by Rhag proteins localized to pillar cells whereas the translocation across the epithelium is probably enabled by Rhbg and Rhcg proteins confined to basolateral and apical membranes, respectively. In this study, we have provided evidence for a similar distribution of Rh proteins in goldfish gill, with some additional sites being identified depending on whether the ILCM was present or absent. In the fish acclimated to 25°C (no ILCM), there was also evidence of Rhag expression within epithelial cells of the lamella and filament. With the appearance of the ILCM in fish acclimated to 7°C, the expression of Rh proteins on pillar cells and lamellar epithelial cells was retained but additionally there was a striking spatial reorganization of all four Rh proteins (but in particular Rhag and Rhbg) to the outer edge of the ILCM. As demonstrated previously (Mitrovic et al., 2009; Mitrovic and Perry, 2009), there was also a profound redistribution of Na⁺/K⁺-ATPase-enriched cells to the outer edge of the ILCM; a subset of these Na⁺/K⁺-ATPase-enriched cells also expressed Rh proteins (e.g. Figs 6, 7, 8, 9). Just as the relocation of Na⁺/K⁺-ATPase-enriched cells to the outer edge of the ILCM may benefit branchial Cl^– uptake (Mitrovic and Perry, 2009), the similar redistribution of Rh proteins may be a mechanism to facilitate the final step of ammonia excretion, its entry into the water. With an ILCM present, the standard localization of Rh proteins would still be required to allow ammonia passage across the lamellar epithelium but its movement throughout the ILCM would require Rh proteins (e.g. Rhbg; Fig. 7) as would its exit into the water. It would be interesting to determine whether the Rh protein-containing cells on the outer edge of the ILCM also express acid transporting proteins such as V-type H⁺-ATPase or Na⁺/H⁺ exchangers (NHE2 or NHE3) to facilitate ammonia excretion by acid trapping (Wright et al., 1989; Shih et al., 2008; Wright and Wood, 2009). Despite the obvious redistribution of Rh proteins in fish acclimated to 7°C, there was no increase in the total quantity of these proteins as determined by western blotting.

The physical covering of the gill lamellae by an ILCM under conditions of reduced O₂ demand can be an effective strategy to reduce the energetic costs of repaying ionic losses across the gill but only if normal physiological functions can be maintained. In the present study, ammonia excretion was maintained in fish possessing covered lamellae probably as a result of the redistribution of ammonia-transporting cells to the outer edges of the ILCM where they can remain in contact with the water. This response is similar to the migration of mitochondria-rich cells to the outer edge of the ILCM that was previously reported by Mitrovic and Perry (Mitrovic and Perry, 2009) and the redistribution of cells expressing mRNA for V-type H⁺-ATPase and NHE3 (Julia Bradshaw and S.F.P., unpublished data). Presumably the redistribution of ion-transporting cells is crucial to maintaining ionic homeostasis when the lamellae are covered. Finally, O₂ sensing, thought to be mediated by neuroepithelial cells within the gill filament and scattered randomly on the lamellae is believed to be sustained in fish experiencing gill remodelling because of a redistribution of neuroepithelial cells to the distal regions of the lamellae (Tzaneva and Perry, 2010). Thus, the transfer of functions from the proximal regions of lamellae to more distal regions or the edge of the ILCM appears to be a common strategy to maintain physiological processes when the gill undergoes remodelling.